Topic

Productivity

Experiments and field studies that measure what AI actually does to output, speed and quality, rather than asking people whether they use it.

Filter in library

Developers expected AI to speed them up. It slowed them down.% change in speed (negative = slower)

Developers' forecast before study

+24%

Developers' belief after study

+20%

Measured effect

−19%

Source: METR, Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity, 2025.

Reports on productivity

Bain · Sep 2026Technology Report 2026

Leaders expect a 95% developer productivity uplift in 1–2 years; current gains are 20–27%. Bain says AI moves bottlenecks into review and coordination.Executive survey

Veracode · Jul 20262026 GenAI Code Security Report

AI models now write syntactically correct code nearly always, but about 44% of generated code still carried a known vulnerability.Benchmark

DX · Jul 2026State of AI Impact in Engineering: Q2 2026 Report

AI now writes 52.7% of code and AI spend rose ~28x in tech, but time on new features stayed flat and change confidence fell 6.1%.Usage data

Glean · Jun 2026Work AI Index 2026: Botsitting, Botshitting & the Hidden Human Labor of AI at Work

Workers report 11 hours a week saved but 6.4 hours spent supervising and fixing AI; only 13% see significant organizational gains.Worker survey

NBER · Feb 2026Firm Data on AI (NBER Working Paper 34836)

Across four countries, most firms use AI, but nine in ten executives report no effect yet on jobs or productivity at their own firm.Executive survey

NBER · Feb 2026Does Generative AI Narrow Education-Based Productivity Gaps? Evidence from a Randomized Experiment

In a randomized test, an AI assistant lifted everyone's scores and cut the gap between higher- and lower-education workers by 75%.Experiment

Anthropic · Nov 2025Estimating AI Productivity Gains from Claude Conversations

Claude's own estimates of 100,000 conversations suggest about 80% time savings per task, before counting time spent checking AI output.Usage data

CAIS · Oct 2025Remote Labor Index: Measuring AI Automation of Remote Work

On 240 paid freelance projects worth over $140,000, the best agent tested produced client-acceptable work 2.5% of the time.Benchmark

OpenAI · Sep 2025GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks

Experts compared AI and human deliverables on real tasks from 44 occupations; the best model matched or beat the expert on 47.6%.Benchmark

Faros AI · Jul 2025The AI Productivity Paradox Report 2025

High-AI teams merged 98% more pull requests, but review time rose 91% and company-level delivery metrics did not improve.Usage data

METR · Jul 2025Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity

In a randomized trial, experienced developers were 19% slower with AI tools, yet afterward believed AI had sped them up by 20%.Experiment

Collections on this topic

17 reportsBeyond usage: what AI actually doesExperiments, payroll data and spending records that measure outcomes, not whether people say they use AI.

Related topics

Workforce Agents ROI & value Strategy Governance & risk